packwise-skills 1.0.0 → 1.2.0

package/sub-skills/desktop/scenarios.md

@@ -1,443 +1,443 @@
-# Desktop Application Scenarios Guide
-
-Covers the most common desktop application types for indie developers and small teams. Each scenario provides specific technology recommendations and packaging considerations.
-
----
-
-## Scenario 1: AI Chatbot / Knowledge Base
-
-**Reference Products**: Claude Desktop, ChatGPT Desktop, OpenCat
-
-**Recommended Stack**: Electron + React + Express + SQLite
-
-**Key Requirements**:
-- LLM API calls (streaming response)
-- Local conversation history persistence
-- TTS voice synthesis (optional)
-- Multi-model switching
-
-**Packaging Considerations**:
-- SSE streaming adaptation (ensure correct forwarding in Electron)
-- Store conversation history in SQLite (more reliable than localStorage)
-- Encrypt API keys (AES-256-CBC)
-- TTS audio queue management (prevent concurrent playback)
-
-**Typical Size**: 130–180MB
-
----
-
-## Scenario 2: Cloud Architecture Monitor
-
-**Reference Products**: Portainer Desktop, Lens, k9s GUI
-
-**Recommended Stack**: Electron/Tauri + React + Go backend
-
-**Key Requirements**:
-- Connect to multiple cloud clusters (K8s, Docker)
-- Real-time status refresh (WebSocket)
-- Resource topology visualization
-- Alert push notifications
-
-**Packaging Considerations**:
-- WebSocket connections need correct proxy in Electron
-- kubeconfig file secure storage
-- TLS certificate handling (self-signed certificates)
-- System notification integration
-
-**Typical Size**: 100–150MB
-
----
-
-## Scenario 3: Local ERP / OA System
-
-**Reference Products**: Kingdee Desktop, Yonyou U8 Client
-
-**Recommended Stack**: Electron + Vue/React + Node/Java + MySQL/SQLite
-
-**Key Requirements**:
-- Complex forms and reports
-- Print module (invoices, contracts)
-- File export (Excel, PDF)
-- Multi-role permissions
-
-**Packaging Considerations**:
-- Print: `window.print()` or system print API
-- Excel export: `xlsx` library (note file save path in Electron)
-- PDF generation: `pdfmake` or `puppeteer`
-- Large data table virtual scrolling
-- Offline-first architecture (local cache, sync on reconnect)
-
-**Typical Size**: 150–200MB
-
----
-
-## Scenario 4: Quantitative Trading Terminal
-
-**Reference Products**: JoinQuant Terminal, QMT, TradingView Desktop
-
-**Recommended Stack**: Electron + React + Python/C++ backend
-
-**Key Requirements**:
-- Real-time market data push (WebSocket/MQTT)
-- K-line chart rendering (Canvas/WebGL)
-- Backtesting engine
-- Strategy editor
-- Order execution
-
-**Packaging Considerations**:
-- High-frequency data: WebSocket connection pool management
-- Chart performance: `lightweight-charts` or `echarts` + Canvas
-- Python backend integration: subprocess or HTTP API with Python backtest engine
-- Security: encrypt trading keys
-- Offline cache: historical market data in SQLite
-
-**Typical Size**: 100–200MB
-
----
-
-## Scenario 5: IoT Control Panel
-
-**Reference Products**: Tuya Smart Desktop, Home Assistant Desktop
-
-**Recommended Stack**: Tauri/Electron + Vue + Go/MQTT
-
-**Key Requirements**:
-- MQTT device communication
-- Device state real-time display
-- Scene automation configuration
-- Data history charts
-
-**Packaging Considerations**:
-- MQTT client: `mqtt.js` (Web) or native MQTT library
-- Device discovery: UDP broadcast (requires native network permission)
-- Low power: reduce refresh rate when backgrounded
-- System tray: resident background, notification push
-
-**Typical Size**: 50–100MB
-
----
-
-## Scenario 6: IM Chat Client
-
-**Reference Products**: Telegram Desktop, Feishu, WeCom
-
-**Recommended Stack**: Electron/Tauri + React + Go/Rust
-
-**Key Requirements**:
-- Real-time messaging (WebSocket)
-- File/image transfer
-- Group/channel support
-- Message search
-- Local message cache
-
-**Packaging Considerations**:
-- Message encryption: E2E encryption (Signal Protocol or custom)
-- File transfer: chunked upload/download, resume support
-- Message storage: SQLite + FTS5 full-text search
-- Notification: native system notifications + unread badge
-- Custom protocol: `myapp://` protocol handler (for link jumps)
-
-**Typical Size**: 100–200MB
-
----
-
-## Scenario 7: SaaS Desktop Client
-
-**Reference Products**: Notion Desktop, Linear, Slack
-
-**Recommended Stack**: Electron/Tauri + React/Vue
-
-**Key Requirements**:
-- Feature parity with web version
-- Desktop notifications
-- Offline support (optional)
-- File drag-and-drop upload
-- Deep links (`myapp://`)
-
-**Packaging Considerations**:
-- If essentially a web wrapper: Pake or Electron both work
-- Offline support: Service Worker + IndexedDB
-- Auto-update: `electron-updater` (Electron) or Tauri built-in updater
-- Single instance lock: prevent duplicate windows
-- Window state memory: remember size and position
-
-**Typical Size**: 100–150MB
-
----
-
-## Scenario 8: Database Management Tool
-
-**Reference Products**: Navicat, DBeaver, pgAdmin Desktop
-
-**Recommended Stack**: Electron + React + Go/Rust
-
-**Key Requirements**:
-- Multi-database connection (MySQL/PostgreSQL/SQLite/MongoDB)
-- SQL editor (syntax highlighting, autocomplete)
-- Table structure management
-- Data import/export
-- Query history
-
-**Packaging Considerations**:
-- Multi-database drivers: Go/Rust backend wraps native drivers, communicates via IPC
-- SQL editor: Monaco Editor or CodeMirror
-- Large dataset handling: virtual scrolling + pagination
-- Connection config encrypted storage
-
-**Typical Size**: 100–200MB
-
----
-
-## Scenario 9: Indie Game
-
-**Reference Products**: Stardew Valley, Hollow Knight, Hades
-
-**Recommended Engine**:
-- **Unity**: 2D/3D universal, C#, most mature cross-platform
-- **Godot**: Free and open-source, GDScript/C#, lightweight
-
-**Key Requirements**:
-- Game loop and rendering
-- Input handling (keyboard/mouse/gamepad)
-- Audio engine
-- Save system
-- Steam/itch.io integration
-
-**Packaging Considerations**:
-- Unity: IL2CPP build + Asset Bundles
-- Godot: export templates + resource packaging
-- Steam integration: Steamworks SDK (achievements, leaderboard, cloud saves)
-- Performance: shader compilation, texture compression, LOD system
-
-**Typical Size**: 200–500MB
-
----
-
-## Scenario 10: Online Education Client
-
-**Reference Products**: Youdao Dictionary, Tencent Classroom, Zoom
-
-**Recommended Stack**: Electron + React + WebRTC
-
-**Key Requirements**:
-- Video live/recorded playback
-- Screen sharing
-- Whiteboard functionality
-- Chat/danmaku
-- Courseware display (PPT/PDF)
-
-**Packaging Considerations**:
-- WebRTC: `simple-peer` or `mediasoup-client`
-- Screen sharing: Electron `desktopCapturer` API
-- Video playback: HLS/DASH streaming (`hls.js`, `dash.js`)
-- Whiteboard: Canvas drawing + real-time sync
-- Audio: echo cancellation (AEC), noise suppression (ANS)
-
-**Typical Size**: 150–250MB
-
----
-
-## Scenario 11: Note-Taking / Knowledge Base
-
-**Reference Products**: Obsidian, Logseq, Notion Desktop
-
-**Recommended Stack**: Electron/Tauri + React + SQLite
-
-**Key Requirements**:
-- Markdown editor
-- Bidirectional linking
-- Full-text search
-- Local file sync
-- Plugin system (optional)
-
-**Packaging Considerations**:
-- Markdown editor: `@codemirror/view` or `tiptap`
-- Full-text search: SQLite FTS5 or `minisearch`
-- File watching: `chokidar` for local file changes
-- Image paste: `clipboard.readImage()` + save locally
-- Performance: virtual scrolling + lazy rendering for large note sets
-
-**Typical Size**: 100–150MB
-
----
-
-## Universal Packaging Checklist
-
-Regardless of scenario, confirm before packaging:
-
-1. Data persistence path correct (userData directory)
-2. API keys encrypted storage
-3. Static resource paths adapted (image embedding or public/)
-4. Native modules correctly packaged (asarUnpack)
-5. Package logo and copyright info correct
-6. Intermediate artifacts cleaned
-7. First install launches successfully
-8. Data clear leaves no residue
-
----
-
-## Scenario 12: SaaS Desktop Client with Payment System
-
-**Reference Products**: Notion, Linear, Figma Desktop, Slack
-
-**Recommended Stack**: Electron/Tauri + React/Vue + Stripe/Alipay
-
-**Key Requirements**:
-- Feature parity with web version
-- Subscription management (monthly/yearly plans)
-- Payment integration (Stripe for global, Alipay/WeChat Pay for China)
-- License verification (prevent piracy)
-- Trial period management
-
-**Packaging Considerations**:
-- Payment: Use Stripe Checkout (redirect to Stripe-hosted page, PCI-compliant)
-- License verification: Server-side validation on startup; cache license locally with expiration
-- Anti-piracy: Bind license to machine ID (`os.hostname()` + hardware fingerprint)
-- Auto-update: electron-updater with differential updates (reduce bandwidth for paying users)
-- Offline grace period: Allow 7-30 days offline before requiring re-validation
-
-**Real Problem & Solution**:
-- **Problem**: Users share license keys. Each key works on unlimited machines.
-- **Solution**: Generate machine-specific activation tokens. Allow 3-5 activations per license. Server-side deactivation API.
-
-**Typical Size**: 100–150MB
-
----
-
-## Scenario 13: Offline-First Field Service Application
-
-**Reference Products**: Fulcrum, ProntoForms, custom industrial inspection tools
-
-**Recommended Stack**: Electron + React + SQLite + Sync Engine
-
-**Key Requirements**:
-- Must work completely offline (factory floor, remote areas)
-- Data sync when connectivity returns (conflict resolution)
-- Photo/video capture with annotation
-- Form builder for custom inspection templates
-- GPS/location tagging
-
-**Packaging Considerations**:
-- Database: SQLite with WAL mode (better concurrent read/write)
-- Sync engine: CRDT (Conflict-free Replicated Data Types) or last-write-wins with conflict UI
-- Media storage: Store photos/videos in userData directory; compress before sync
-- Network detection: `navigator.onLine` + actual connectivity check (ping server)
-- Large data: Paginated sync; don't try to sync 10GB of photos at once
-
-**Real Problem & Solution**:
-- **Problem**: User collects 500 inspection records offline. When reconnecting, sync takes 20 minutes and fails midway.
-- **Solution**: Incremental sync with checkpoint. Each record synced individually. Resume from last successful record on failure. Show progress bar.
-
-**Typical Size**: 150–250MB
-
----
-
-## Scenario 14: AI-Powered Desktop Application with Local Model
-
-**Reference Products**: Ollama Desktop, LM Studio, Cursor, Continue.dev
-
-**Recommended Stack**: Electron + React + Ollama/llama.cpp + SQLite
-
-**Key Requirements**:
-- Local LLM inference (no cloud API dependency)
-- Model management (download, switch, delete models)
-- Chat interface with streaming response
-- Document RAG (Retrieval-Augmented Generation)
-- GPU detection and optimization
-
-**Packaging Considerations**:
-- Model storage: Large files (4-70GB); store outside app directory; allow user to choose location
-- GPU detection: Check CUDA/Metal/Vulkan availability at startup; fallback to CPU
-- Memory management: Monitor RAM/VRAM usage; warn before loading models that exceed available memory
-- Streaming: Use Server-Sent Events (SSE) between main process and renderer
-- Binary distribution: Ship llama.cpp binary as extraResource; detect platform at runtime
-
-**Real Problem & Solution**:
-- **Problem**: App ships with llama.cpp binary compiled for AVX2. User's CPU only supports AVX. App crashes silently.
-- **Solution**: Detect CPU features at startup. Ship multiple binaries (AVX, AVX2, ARM NEON). Select the best one dynamically. Show clear error if none compatible.
-
-**Typical Size**: 150–200MB (without models)
-
----
-
-## Scenario 15: Hardware Integration Application (Serial Port / USB / BLE)
-
-**Reference Products**: Arduino IDE, PlatformIO, PuTTY, serial terminal tools
-
-**Recommended Stack**: Electron + React + serialport/noble/usb + native modules
-
-**Key Requirements**:
-- Serial port communication (RS-232, RS-485, USB-to-serial)
-- USB device access (HID, custom protocols)
-- Bluetooth Low Energy (BLE) communication
-- Real-time data visualization (oscilloscope-style charts)
-- Firmware flashing / OTA update for connected devices
-
-**Packaging Considerations**:
-- Native modules: `serialport`, `usb`, `@abandonware/noble` — ALL require `asarUnpack`
-- Permissions: macOS requires entitlements for USB/BLE access; Windows may need driver installation
-- Platform differences: Serial port paths differ (`/dev/ttyUSB0` vs `COM3`)
-- Real-time data: Use `requestAnimationFrame` for chart rendering; don't block main process
-- Firmware flashing: Binary protocol over serial; implement retry + verification
-
-**Real Problem & Solution**:
-- **Problem**: `serialport` module works in dev but fails in packaged app with "Cannot find module".
-- **Solution**: `asarUnpack` + `electron-rebuild`. The `.node` binary must be rebuilt for Electron's ABI and placed in unpacked directory.
-
-**Typical Size**: 130–180MB
-
----
-
-## Scenario 16: Multi-Tenant Enterprise Application
-
-**Reference Products**: Odoo Desktop, SAP Business One, custom ERP clients
-
-**Recommended Stack**: Electron + React/Vue + Go/Rust backend + PostgreSQL
-
-**Key Requirements**:
-- Multi-tenant data isolation (separate databases or schemas per tenant)
-- Role-based access control (RBAC)
-- Audit logging (who did what when)
-- Single Sign-On (SSO) via SAML/OIDC
-- Data export (Excel, PDF, CSV)
-
-**Packaging Considerations**:
-- Database: Don't bundle PostgreSQL; connect to external server. SQLite only for local cache.
-- SSO: Open external browser for OAuth flow; handle callback via custom protocol (`myapp://callback`)
-- Audit log: Write to separate SQLite database; don't mix with business data
-- Multi-tenant switching: Store tenant token in keychain (not localStorage)
-- Large datasets: Virtual scrolling for tables with 10,000+ rows
-
-**Real Problem & Solution**:
-- **Problem**: User switches between 5 tenants. Each tenant has different permissions. App shows stale permissions from previous tenant.
-- **Solution**: Clear all cached permissions on tenant switch. Re-fetch user role from server. Show loading state during transition.
-
-**Typical Size**: 150–200MB
-
----
-
-## Scenario 17: Real-Time Collaboration Application
-
-**Reference Products**: Figma, Excalidraw, Miro, Notion (collaborative editing)
-
-**Recommended Stack**: Electron/Tauri + React + Yjs/CRDT + WebSocket
-
-**Key Requirements**:
-- Real-time collaborative editing (multiple cursors, live changes)
-- Conflict resolution (CRDT or OT)
-- Presence awareness (who's online, where they're editing)
-- Offline editing with sync on reconnect
-- Large document handling
-
-**Packaging Considerations**:
-- CRDT library: Yjs (recommended) or Automerge for conflict-free merging
-- WebSocket: Reconnect with exponential backoff; queue changes while offline
-- Cursor sync: Throttle cursor position updates to 10/second max
-- Performance: Use Web Workers for CRDT operations on large documents
-- Memory: Yjs documents can grow large; implement garbage collection for old operations
-
-**Real Problem & Solution**:
-- **Problem**: 10 users editing a 500-page document simultaneously. Memory usage grows to 2GB+. App becomes unresponsive.
-- **Solution**: Implement document chunking (load only visible pages). Use Yjs sub-documents for lazy loading. Compress old CRDT operations periodically.
-
-**Typical Size**: 100–150MB
+# Desktop Application Scenarios Guide
+
+Covers the most common desktop application types for indie developers and small teams. Each scenario provides specific technology recommendations and packaging considerations.
+
+---
+
+## Scenario 1: AI Chatbot / Knowledge Base
+
+**Reference Products**: Claude Desktop, ChatGPT Desktop, OpenCat
+
+**Recommended Stack**: Electron + React + Express + SQLite
+
+**Key Requirements**:
+- LLM API calls (streaming response)
+- Local conversation history persistence
+- TTS voice synthesis (optional)
+- Multi-model switching
+
+**Packaging Considerations**:
+- SSE streaming adaptation (ensure correct forwarding in Electron)
+- Store conversation history in SQLite (more reliable than localStorage)
+- Encrypt API keys (AES-256-CBC)
+- TTS audio queue management (prevent concurrent playback)
+
+**Typical Size**: 130–180MB
+
+---
+
+## Scenario 2: Cloud Architecture Monitor
+
+**Reference Products**: Portainer Desktop, Lens, k9s GUI
+
+**Recommended Stack**: Electron/Tauri + React + Go backend
+
+**Key Requirements**:
+- Connect to multiple cloud clusters (K8s, Docker)
+- Real-time status refresh (WebSocket)
+- Resource topology visualization
+- Alert push notifications
+
+**Packaging Considerations**:
+- WebSocket connections need correct proxy in Electron
+- kubeconfig file secure storage
+- TLS certificate handling (self-signed certificates)
+- System notification integration
+
+**Typical Size**: 100–150MB
+
+---
+
+## Scenario 3: Local ERP / OA System
+
+**Reference Products**: Kingdee Desktop, Yonyou U8 Client
+
+**Recommended Stack**: Electron + Vue/React + Node/Java + MySQL/SQLite
+
+**Key Requirements**:
+- Complex forms and reports
+- Print module (invoices, contracts)
+- File export (Excel, PDF)
+- Multi-role permissions
+
+**Packaging Considerations**:
+- Print: `window.print()` or system print API
+- Excel export: `xlsx` library (note file save path in Electron)
+- PDF generation: `pdfmake` or `puppeteer`
+- Large data table virtual scrolling
+- Offline-first architecture (local cache, sync on reconnect)
+
+**Typical Size**: 150–200MB
+
+---
+
+## Scenario 4: Quantitative Trading Terminal
+
+**Reference Products**: JoinQuant Terminal, QMT, TradingView Desktop
+
+**Recommended Stack**: Electron + React + Python/C++ backend
+
+**Key Requirements**:
+- Real-time market data push (WebSocket/MQTT)
+- K-line chart rendering (Canvas/WebGL)
+- Backtesting engine
+- Strategy editor
+- Order execution
+
+**Packaging Considerations**:
+- High-frequency data: WebSocket connection pool management
+- Chart performance: `lightweight-charts` or `echarts` + Canvas
+- Python backend integration: subprocess or HTTP API with Python backtest engine
+- Security: encrypt trading keys
+- Offline cache: historical market data in SQLite
+
+**Typical Size**: 100–200MB
+
+---
+
+## Scenario 5: IoT Control Panel
+
+**Reference Products**: Tuya Smart Desktop, Home Assistant Desktop
+
+**Recommended Stack**: Tauri/Electron + Vue + Go/MQTT
+
+**Key Requirements**:
+- MQTT device communication
+- Device state real-time display
+- Scene automation configuration
+- Data history charts
+
+**Packaging Considerations**:
+- MQTT client: `mqtt.js` (Web) or native MQTT library
+- Device discovery: UDP broadcast (requires native network permission)
+- Low power: reduce refresh rate when backgrounded
+- System tray: resident background, notification push
+
+**Typical Size**: 50–100MB
+
+---
+
+## Scenario 6: IM Chat Client
+
+**Reference Products**: Telegram Desktop, Feishu, WeCom
+
+**Recommended Stack**: Electron/Tauri + React + Go/Rust
+
+**Key Requirements**:
+- Real-time messaging (WebSocket)
+- File/image transfer
+- Group/channel support
+- Message search
+- Local message cache
+
+**Packaging Considerations**:
+- Message encryption: E2E encryption (Signal Protocol or custom)
+- File transfer: chunked upload/download, resume support
+- Message storage: SQLite + FTS5 full-text search
+- Notification: native system notifications + unread badge
+- Custom protocol: `myapp://` protocol handler (for link jumps)
+
+**Typical Size**: 100–200MB
+
+---
+
+## Scenario 7: SaaS Desktop Client
+
+**Reference Products**: Notion Desktop, Linear, Slack
+
+**Recommended Stack**: Electron/Tauri + React/Vue
+
+**Key Requirements**:
+- Feature parity with web version
+- Desktop notifications
+- Offline support (optional)
+- File drag-and-drop upload
+- Deep links (`myapp://`)
+
+**Packaging Considerations**:
+- If essentially a web wrapper: Pake or Electron both work
+- Offline support: Service Worker + IndexedDB
+- Auto-update: `electron-updater` (Electron) or Tauri built-in updater
+- Single instance lock: prevent duplicate windows
+- Window state memory: remember size and position
+
+**Typical Size**: 100–150MB
+
+---
+
+## Scenario 8: Database Management Tool
+
+**Reference Products**: Navicat, DBeaver, pgAdmin Desktop
+
+**Recommended Stack**: Electron + React + Go/Rust
+
+**Key Requirements**:
+- Multi-database connection (MySQL/PostgreSQL/SQLite/MongoDB)
+- SQL editor (syntax highlighting, autocomplete)
+- Table structure management
+- Data import/export
+- Query history
+
+**Packaging Considerations**:
+- Multi-database drivers: Go/Rust backend wraps native drivers, communicates via IPC
+- SQL editor: Monaco Editor or CodeMirror
+- Large dataset handling: virtual scrolling + pagination
+- Connection config encrypted storage
+
+**Typical Size**: 100–200MB
+
+---
+
+## Scenario 9: Indie Game
+
+**Reference Products**: Stardew Valley, Hollow Knight, Hades
+
+**Recommended Engine**:
+- **Unity**: 2D/3D universal, C#, most mature cross-platform
+- **Godot**: Free and open-source, GDScript/C#, lightweight
+
+**Key Requirements**:
+- Game loop and rendering
+- Input handling (keyboard/mouse/gamepad)
+- Audio engine
+- Save system
+- Steam/itch.io integration
+
+**Packaging Considerations**:
+- Unity: IL2CPP build + Asset Bundles
+- Godot: export templates + resource packaging
+- Steam integration: Steamworks SDK (achievements, leaderboard, cloud saves)
+- Performance: shader compilation, texture compression, LOD system
+
+**Typical Size**: 200–500MB
+
+---
+
+## Scenario 10: Online Education Client
+
+**Reference Products**: Youdao Dictionary, Tencent Classroom, Zoom
+
+**Recommended Stack**: Electron + React + WebRTC
+
+**Key Requirements**:
+- Video live/recorded playback
+- Screen sharing
+- Whiteboard functionality
+- Chat/danmaku
+- Courseware display (PPT/PDF)
+
+**Packaging Considerations**:
+- WebRTC: `simple-peer` or `mediasoup-client`
+- Screen sharing: Electron `desktopCapturer` API
+- Video playback: HLS/DASH streaming (`hls.js`, `dash.js`)
+- Whiteboard: Canvas drawing + real-time sync
+- Audio: echo cancellation (AEC), noise suppression (ANS)
+
+**Typical Size**: 150–250MB
+
+---
+
+## Scenario 11: Note-Taking / Knowledge Base
+
+**Reference Products**: Obsidian, Logseq, Notion Desktop
+
+**Recommended Stack**: Electron/Tauri + React + SQLite
+
+**Key Requirements**:
+- Markdown editor
+- Bidirectional linking
+- Full-text search
+- Local file sync
+- Plugin system (optional)
+
+**Packaging Considerations**:
+- Markdown editor: `@codemirror/view` or `tiptap`
+- Full-text search: SQLite FTS5 or `minisearch`
+- File watching: `chokidar` for local file changes
+- Image paste: `clipboard.readImage()` + save locally
+- Performance: virtual scrolling + lazy rendering for large note sets
+
+**Typical Size**: 100–150MB
+
+---
+
+## Universal Packaging Checklist
+
+Regardless of scenario, confirm before packaging:
+
+1. Data persistence path correct (userData directory)
+2. API keys encrypted storage
+3. Static resource paths adapted (image embedding or public/)
+4. Native modules correctly packaged (asarUnpack)
+5. Package logo and copyright info correct
+6. Intermediate artifacts cleaned
+7. First install launches successfully
+8. Data clear leaves no residue
+
+---
+
+## Scenario 12: SaaS Desktop Client with Payment System
+
+**Reference Products**: Notion, Linear, Figma Desktop, Slack
+
+**Recommended Stack**: Electron/Tauri + React/Vue + Stripe/Alipay
+
+**Key Requirements**:
+- Feature parity with web version
+- Subscription management (monthly/yearly plans)
+- Payment integration (Stripe for global, Alipay/WeChat Pay for China)
+- License verification (prevent piracy)
+- Trial period management
+
+**Packaging Considerations**:
+- Payment: Use Stripe Checkout (redirect to Stripe-hosted page, PCI-compliant)
+- License verification: Server-side validation on startup; cache license locally with expiration
+- Anti-piracy: Bind license to machine ID (`os.hostname()` + hardware fingerprint)
+- Auto-update: electron-updater with differential updates (reduce bandwidth for paying users)
+- Offline grace period: Allow 7-30 days offline before requiring re-validation
+
+**Real Problem & Solution**:
+- **Problem**: Users share license keys. Each key works on unlimited machines.
+- **Solution**: Generate machine-specific activation tokens. Allow 3-5 activations per license. Server-side deactivation API.
+
+**Typical Size**: 100–150MB
+
+---
+
+## Scenario 13: Offline-First Field Service Application
+
+**Reference Products**: Fulcrum, ProntoForms, custom industrial inspection tools
+
+**Recommended Stack**: Electron + React + SQLite + Sync Engine
+
+**Key Requirements**:
+- Must work completely offline (factory floor, remote areas)
+- Data sync when connectivity returns (conflict resolution)
+- Photo/video capture with annotation
+- Form builder for custom inspection templates
+- GPS/location tagging
+
+**Packaging Considerations**:
+- Database: SQLite with WAL mode (better concurrent read/write)
+- Sync engine: CRDT (Conflict-free Replicated Data Types) or last-write-wins with conflict UI
+- Media storage: Store photos/videos in userData directory; compress before sync
+- Network detection: `navigator.onLine` + actual connectivity check (ping server)
+- Large data: Paginated sync; don't try to sync 10GB of photos at once
+
+**Real Problem & Solution**:
+- **Problem**: User collects 500 inspection records offline. When reconnecting, sync takes 20 minutes and fails midway.
+- **Solution**: Incremental sync with checkpoint. Each record synced individually. Resume from last successful record on failure. Show progress bar.
+
+**Typical Size**: 150–250MB
+
+---
+
+## Scenario 14: AI-Powered Desktop Application with Local Model
+
+**Reference Products**: Ollama Desktop, LM Studio, Cursor, Continue.dev
+
+**Recommended Stack**: Electron + React + Ollama/llama.cpp + SQLite
+
+**Key Requirements**:
+- Local LLM inference (no cloud API dependency)
+- Model management (download, switch, delete models)
+- Chat interface with streaming response
+- Document RAG (Retrieval-Augmented Generation)
+- GPU detection and optimization
+
+**Packaging Considerations**:
+- Model storage: Large files (4-70GB); store outside app directory; allow user to choose location
+- GPU detection: Check CUDA/Metal/Vulkan availability at startup; fallback to CPU
+- Memory management: Monitor RAM/VRAM usage; warn before loading models that exceed available memory
+- Streaming: Use Server-Sent Events (SSE) between main process and renderer
+- Binary distribution: Ship llama.cpp binary as extraResource; detect platform at runtime
+
+**Real Problem & Solution**:
+- **Problem**: App ships with llama.cpp binary compiled for AVX2. User's CPU only supports AVX. App crashes silently.
+- **Solution**: Detect CPU features at startup. Ship multiple binaries (AVX, AVX2, ARM NEON). Select the best one dynamically. Show clear error if none compatible.
+
+**Typical Size**: 150–200MB (without models)
+
+---
+
+## Scenario 15: Hardware Integration Application (Serial Port / USB / BLE)
+
+**Reference Products**: Arduino IDE, PlatformIO, PuTTY, serial terminal tools
+
+**Recommended Stack**: Electron + React + serialport/noble/usb + native modules
+
+**Key Requirements**:
+- Serial port communication (RS-232, RS-485, USB-to-serial)
+- USB device access (HID, custom protocols)
+- Bluetooth Low Energy (BLE) communication
+- Real-time data visualization (oscilloscope-style charts)
+- Firmware flashing / OTA update for connected devices
+
+**Packaging Considerations**:
+- Native modules: `serialport`, `usb`, `@abandonware/noble` — ALL require `asarUnpack`
+- Permissions: macOS requires entitlements for USB/BLE access; Windows may need driver installation
+- Platform differences: Serial port paths differ (`/dev/ttyUSB0` vs `COM3`)
+- Real-time data: Use `requestAnimationFrame` for chart rendering; don't block main process
+- Firmware flashing: Binary protocol over serial; implement retry + verification
+
+**Real Problem & Solution**:
+- **Problem**: `serialport` module works in dev but fails in packaged app with "Cannot find module".
+- **Solution**: `asarUnpack` + `electron-rebuild`. The `.node` binary must be rebuilt for Electron's ABI and placed in unpacked directory.
+
+**Typical Size**: 130–180MB
+
+---
+
+## Scenario 16: Multi-Tenant Enterprise Application
+
+**Reference Products**: Odoo Desktop, SAP Business One, custom ERP clients
+
+**Recommended Stack**: Electron + React/Vue + Go/Rust backend + PostgreSQL
+
+**Key Requirements**:
+- Multi-tenant data isolation (separate databases or schemas per tenant)
+- Role-based access control (RBAC)
+- Audit logging (who did what when)
+- Single Sign-On (SSO) via SAML/OIDC
+- Data export (Excel, PDF, CSV)
+
+**Packaging Considerations**:
+- Database: Don't bundle PostgreSQL; connect to external server. SQLite only for local cache.
+- SSO: Open external browser for OAuth flow; handle callback via custom protocol (`myapp://callback`)
+- Audit log: Write to separate SQLite database; don't mix with business data
+- Multi-tenant switching: Store tenant token in keychain (not localStorage)
+- Large datasets: Virtual scrolling for tables with 10,000+ rows
+
+**Real Problem & Solution**:
+- **Problem**: User switches between 5 tenants. Each tenant has different permissions. App shows stale permissions from previous tenant.
+- **Solution**: Clear all cached permissions on tenant switch. Re-fetch user role from server. Show loading state during transition.
+
+**Typical Size**: 150–200MB
+
+---
+
+## Scenario 17: Real-Time Collaboration Application
+
+**Reference Products**: Figma, Excalidraw, Miro, Notion (collaborative editing)
+
+**Recommended Stack**: Electron/Tauri + React + Yjs/CRDT + WebSocket
+
+**Key Requirements**:
+- Real-time collaborative editing (multiple cursors, live changes)
+- Conflict resolution (CRDT or OT)
+- Presence awareness (who's online, where they're editing)
+- Offline editing with sync on reconnect
+- Large document handling
+
+**Packaging Considerations**:
+- CRDT library: Yjs (recommended) or Automerge for conflict-free merging
+- WebSocket: Reconnect with exponential backoff; queue changes while offline
+- Cursor sync: Throttle cursor position updates to 10/second max
+- Performance: Use Web Workers for CRDT operations on large documents
+- Memory: Yjs documents can grow large; implement garbage collection for old operations
+
+**Real Problem & Solution**:
+- **Problem**: 10 users editing a 500-page document simultaneously. Memory usage grows to 2GB+. App becomes unresponsive.
+- **Solution**: Implement document chunking (load only visible pages). Use Yjs sub-documents for lazy loading. Compress old CRDT operations periodically.
+
+**Typical Size**: 100–150MB